The present invention relates to map inference technology, and more specifically, to a method for generating a road network and an apparatus thereof.
Recent years, the technology for road map inference using Global Position System (GPS) has attracted a lot of research interests, since in such technology, the road map is inferred from a significant amount of trajectory data of actual route traveled by vehicles obtained from GPS. In comparison to the traditional geographical survey approach, it is more efficient to infer the road map from GPS. In many applications, for example, trajectory pattern mining, travel activity analysis, vehicle fleet management, and etc., the inferred road map can be typically used to transform raw trajectories to abstract trajectories with reduced data size and/or enriched semantics.
A number of technologies have been developed to infer road maps using a great amount of trajectory data of actual routes traveled by vehicles obtained from GPS. Unfortunately, in the prior art, the generated road maps are routable fully functional road maps that can be used for navigation purpose. On one hand, individual GPS trajectories' geometrical features need to be computed for such fully functional road maps, which stands for significant computation overhead and is hardly scalable. For example, a map inference computation from about 1,000 GPS trajectories will take approximately one hour. On the other hand, traffic information for navigation contained in the fully functional road map is not necessary for applications other than navigation related applications, such as trajectory pattern mining, travel activity analysis, vehicle fleet management, and etc. Thus, the fully functional road map is redundant for these applications. Instead, maps with reduced data size can be used to facilitate transmission and storage.
Therefore, a technology is needed for generating a road map from raw GPS trajectory data in a more efficient and/or scalable manner.